✧✦Catherine✦✧remember i bought a cheap Chinese motherboard with a Xeon included in the price?
i just ran lspci on it.
look at it. what is even going on
biblically accurate PCI bus
full text if you're curious https://paste.debian.net/hidden/a5c1fead
equinox@whitequark this is sadly normal, everything on bus ff: is Intel having engorged body parts for unnecessary PCI devices 
(To be fair, "look good" is not a criteria for PCI bus/interface design)
@equinox pretty sure it's a debug interface that's not meant to be exposed on motherboards shipped to customers
Andrew Zonenberg@whitequark @equinox this is one of my recent xeon workstations https://paste.debian.net/hidden/2b8d5ed8
doesnt look all that different?
@whitequark @azonenberg it's not like any of this exists as actual PCI bus or devices, they just have a "PCI-looking" configuration interface bus in the CPU and aren't afraid to use it…
In some cases you actually get drivers that work for the same component across multiple CPU generations, while other things around it change… in other cases [it feels like] they just threw shit at the wall and some stuck
@whitequark @equinox there's almost certainly a ton of fun registers for memory training etc in there if you know what to look for
AlexB@azonenberg @whitequark @equinox it's pretty progressive actually, instead of copy pasting a magic base address from (hopefully) a manual (or figure out where the BIOS has put it), you get normal pci id-based enumeration and mmio to a BAR (or config space)
@azonenberg @whitequark @equinox of course, some of those are sometimes documented like, "this looks like PCI for cosmetic reasons. If you actually try to map the BAR somewhere else you crash the machine". Then the BIOS of course hides the device entirely as a workaround
Jernej Simončič �@azonenberg @whitequark @equinox Yeah, it's pretty normal for a Xeon board. Here's lspci from a random SuperMicro: https://clipboards.drifting-through.space/a4f81ed6/
NatTheCathttps://paste.debian.net/hidden/cf6f25fc same here lol, just two of them bc dual socket board(and other vendor-specific junk with the exception of like, one device which appears as a usb 3.1 host controller, 08:00.0, PCIE device I added, the rest is just the machine.)
@whitequark I don't think so, my 'proper' Supermicro system looks no different really
(I might be missing one or two things, but the overall picture is really the same.)
@whitequark https://paste.debian.net/hidden/586e5aec dual socket Xeon E5 v2, allows easy distinguishing whats's CPU and what's chipset
EveyPN@whitequark @equinox Na this is just how they expose the uncore performance counters. Quite cool how much you can debug/ trace with those. https://builders.intel.com/docs/networkbuilders/4th-gen-intel-xeon-scalable-processor-xcc-codename-sapphire-rapids-uncore-performance-1746785301.pdf for example for a somewhat modern platform